Introduction

Antimicrobial treatment and prevention strategies have been plagued by bacterial resistance arising virtually from the onset of their introduction into the health care system. In particular, many strains of the gram-positive bacterium Staphylococcus aureus have acquired broad-spectrum resistance to many traditional therapies including β-lactams1 and last-resort antibiotics such as vancomycin2. These resistant strains of S. aureus are a major public health problem causing severe disease and death3. Additional staphylococcal species including Staphylococcus epidermidis, Staphylococcus lugdunensis, and Staphylococcus haemolyticus cause opportunistic and chronic disease highlighting the need for novel treatment options to combat gram-positive bacterial infections4. Targeting of essential virulence factors in these pathogens has shown promise recently for the development of new therapeutics and vaccines5.

A wide variety of virulence factors that manipulate and modulate host and competing bacterial cells for survival are involved in staphylococcal infections including the phenol soluble modulins (PSMs)4,6,7. PSMs are amphipathic, α-helical peptides that are crucial in various modes of pathogenesis and show potential value as a therapeutic target. While some PSMs have membrane-destructive properties and cytolyze erythrocytes, leukocytes, and other bacteria, others are implicated in biofilm development and detachment, as well as the triggering of receptor-mediated inflammatory responses during infection8,9. PSMα peptides are 20-26 residues with amphipathic helical properties10. They are the predominant type produced by S. aureus and are critical for the majority of virulence roles attributed to PSMs4,11. PSM transport has also been associated with increased resistance to host-produced antimicrobial peptides that share similar amphipathic, α-helical, and membrane-disruptive properties, indicating a possible role for PSM transport systems in subversion of the innate immune system through the efflux of these host defense peptides12.

A PSM transport operon (Pmt) was identified and found to be essential for S. aureus growth given the intrinsic membranolytic properties of PSMs13,14. The Pmt operon encodes two ATP-binding cassette (ABC) transporters known as PmtAB and PmtCD13,15. We previously demonstrated that PmtCD is the minimal Pmt unit necessary for PSMα export and reported a 4 Å cryo-EM structure of S. aureus PmtCD complexed to the hydrolysis-impaired substrate analog adenosine 5′-[γ-thio]triphosphate (ATPγS) and a low resolution (~8 Å) nucleotide-free reconstruction15. PmtCD is a type V ABC transporter16,17 with two PmtD transmembrane domains (TMDs) and two cytosolic protomers of PmtC15. The latter is comprised of a nucleotide-binding domain (NBD) with an ACT (aspartate kinase, chorismate mutase, and prephenate dehydrogenase TyrA) domain insertion that forms a cytosolic basket-like structure critical to PmtCD function and dimerization (Fig. S1)15. Despite the progress that has been made in understanding Pmt systems, the prior low resolution of the apo state (also unpredicted by AI methods) and lack of the ADP product state has hindered the ability to confidently discern and support the unique structural differences that allow for PSM binding and ATP-dependent export.

Here, we present structures of S. aureus PmtCD in the nucleotide-free state reconstituted in three membrane mimetics as well as a structure in complex with ADP. This analysis shows that, regardless of mimetic type, PmtCD adopts a dramatically open conformation in absence of nucleotide with no interaction between its transmembrane domains. Upon nucleotide binding, requisite close packing of the NBDs and in parallel, the TMDs causes the collapse of the large lumen observed in the apo state. These conformations of PmtCD support the use of a lateral access and extrusion mechanism for PSMα export.

Results

Apo PmtCD adopts a large intramembrane lumen in three different membrane mimetics

We have determined the single particle cryo-EM structures of nucleotide-free PmtCD reconstituted into three membrane mimetics: peptidisc, nanodisc or detergent lauryl maltose neopentyl glycol (LMNG) at micellar concentrations to 4.3 Å, 4.3 Å and 3.8 Å resolution, respectively for the unsymmetrized (C1) reconstructions (Figs. 1a and S2–4, and Table 1). In each case, PmtCD adopts a Y-shaped structure with no interaction between the two PmtD TMDs and a large intervening intramembrane lumen. Dimerization is mediated solely by the PmtC NBDs and the asymmetrically disposed PmtC ACT cytosolic domains. Although the overall symmetry of the complex is C1, C2 symmetry averaging improved the local resolution of the TMDs and NBDs and was used to assist model building of these regions.

Fig. 1: Cryo-EM structures of apo PmtCD in three distinct membrane mimetics.
figure 1

a Cryo-EM reconstructions of apo PmtCD reconstituted in different membrane mimetics, from left to right: detergent LMNG, nanodisc, and peptidisc. The region of the map corresponding to PmtC protomers (nucleotide binding domains) are shown in yellow and orange, and PmtD protomers (transmembrane domains) are shown in blue and red. Predicted membrane span and the PmtCD interfaces therein are indicated between the dashed lines. b In all mimetics, apo PmtCD shows the relatively unusual disposition of the two PmtD promoters completely separated, forming a significant intramembrane lumen which varies somewhat in size. PmtCD model shown as cylinders and coloured as in (a). The intramembrane lumen is denoted by the transparent cyan surface. The lateral entrance and internal cavity dimensions, as well as total volumes, are labelled.

Full size image
Table 1 Cryo-EM data collection and refinement statistics
Full size table

The individual PmtD protomer structures characterized in these apo forms are conserved amongst themselves (range of Cα RMSD 0.3–0.4 Å across all 247 residues) and with the prior ATPγS-bound structure (PDB 6xji) (Cα RMSD 1.2–1.3 Å across all 247 residues). The lack of significant conformational differences between open and closed states is suggestive of largely rigid body motions dominating during the ATP driven extrusion of PSM substrate. The TMD architecture, shared amongst type V ABC transporters16,17, features six transmembrane helices (TM1-TM6), an amphipathic connecting helix (CnH) and a coupling helix (CpH) for NBD binding. Two reentrant helices (EH1 and EH2) cap the TMD at the extracellular leaflet of the lipid bilayer (Fig. 2b, c and S1). The PmtC NBD protomer is also structurally similar to that of the ATPγS-bound state (Cα RMSD with closed state ~0.8 Å across all 205 residues excluding the ACT domain) and adopts a typical RecA-like ATPase fold (with a six-stranded β-sheet surrounded by four α-helices) supplemented with a three-stranded antiparallel β-sheet (ABCβ) and an α-helical subdomain (ABCα)17 (Fig. 2b, c, S1 and S5a). The two NBDs homodimerize with a C2 symmetry axis perpendicular to the membrane. An insertion at the cytosolic most C-terminal region of each NBD forms the basket domain that adopts the βαββαβ ACT fold commonly used for binding of small molecules and regulation of metabolic enzyme activities18 (Fig. 2b, c, S1 and S5b). Each of the two ACT domains of PmtC precede a pinning helix (PH) that packs against the NBD-NBD interface (β3, β8, β9, and α6 of one NBD and α5 of the opposing NBD). The ACT domains also homodimerize and the dimer consistently tilts ~25ᵒ to the membrane plane with one protomer packing against its C-terminal PH and the neighbouring protomer orienting away from its PH.

Fig. 2: Cryo-EM structures of PmtCD in peptidisc.
figure 2

a Cryo-EM reconstructions of nucleotide-free PmtCD, ATPγS-bound PmtCD (PDB ID: 6XJI), and ADP-bound PmtCD in peptidisc. PmtC protomers are shown in yellow and orange, and PmtD protomers are shown in blue and red. Predicted membrane interfaces are indicated between the dashed lines. b Cartoon representations of nucleotide-free PmtCD in peptidisc. The exporter is a heterotetramer comprised of two transmembrane domains (TMDs), two nucleotide binding domains (NBDs), and two basket domains. Transmembrane helices are labeled along with the pinning helix (PH), the two reentrant helices (EH1 and EH2), and the connecting helix (CnH) and coupling helix (CpH) for NBD binding. c Cartoon representations of ADP-bound PmtCD in peptidisc. ADP shown in green. Heteroatoms are coloured by type (O, red; P, orange; N, blue).

Full size image

The open state is consistent in overall dimensions with our previous low resolution (~8 Å) cryo-EM reconstruction of the apo form and starkly contrasts with the high-resolution closed state structure of PmtCD in complex with ATPγS15, where the TMDs are in close association more typical of other ABC transporter structures. Importantly here, the expansive open inner lumen specific to the apo form was consistently observed in three different membrane mimetics suggesting it is not an artifact of the solubilization method used (Fig. 1a). The lumen dimensions and volume vary slightly in detailed comparisons of the nucleotide-free PmtCD structures in the distinct membrane mimetics (Fig. 1b). The lateral entrance to the lumen is framed by TM1 and TM5 from opposite protomers along the span of both the inner and outer leaflets of the membrane. The width is approximately consistent along the membrane span and is narrowest in the reconstituted Membrane Scaffold Protein (MSP) nanodisc structure (~12 Å; average distance between Cα atoms of residues projecting into the lumen along helix TM1 and TM5 of opposing PmtD protomers) with LMNG detergent (~16 Å) and the noncovalently assembled peptidisc (~17 Å) being wider. In parallel, the interior of the lumen widens to ~21 Å between TM2 in the nanodisc structure, ~25 Å in the LMNG structure and ~26 Å in the peptidisc structure (see Table S1 for distances between symmetry related residues). The channel volumes are consistent with these at 10045 Å3, 10859 Å3 and 11659 Å3 respectively, as calculated using the 3 V server19. Interestingly, the lumen was smaller in the constrained MSP1D1 nanodiscs (estimated diameter of ∼9.7 nm20) where PmtCD was positioned close to the internal boundary of the nanodisc. We propose the curvature of the MSP protein belt could constrict the relative disposition of the two PmtD TMDs (Fig. S6). These observed differences also suggest a plasticity that could be influenced by membrane dynamics, interactions with other membrane proteins or substrates. Regardless, the wide inner lumen is consistent amongst all membrane mimetics, and of a size that unambiguously precludes direct association of the two TMD domains in the nucleotide-free state.

Two cytosolic entrances to the central lumen are observed, bounded by the inner leaflet of the membrane and the two TMD-NBD modules (Fig. 3). Presumed to be an access point for newly synthesized PSMs from the cytosol, conserved polar residues of the TMDs (S13, S17, K72, T75, R76, S79, and Q80) are suitably located at this proposed substrate entrance to mediate their binding (Fig. 3b, c and S7). The electrostatic surface potential of the exporter reveals a highly positively charged patch at the putative cytosol-membrane interface (Fig. 3d). This is likely used to stabilize the transporter in the membrane through electrostatic interactions with negatively charged phospholipid head groups. Intriguingly, a cluster of lysine residues (K78, K114, K115, K116, K118, and K119) is present on the surface of the NBD adjacent to the nucleotide-binding site (Fig. 3b and S8). The tethering of negatively charged lipid head groups to this extended basic patch of residues on PmtCD would partially enclose the intramembrane lumen to perhaps narrow the trajectory of peptides entering and traversing the exporter.

Fig. 3: Structural features of nucleotide-free PmtCD in peptidisc.
figure 3

a Cavities and lumens of nucleotide-free PmtCD in peptidisc are shown in cyan with PmtC protomers in yellow and orange and PmtD protomers in blue and red. The predicted membrane interfacial regions are indicated by gray disks. b The cytosol-membrane interface and the intracellular regions of the two TMD-NBD modules outline a cytosolic cavity that can accommodate PSMs. The cavity is adjacent to a cluster of positively charged residues (blue) on the NBD and a patch of highly conserved polar residues at the cytosol-membrane interface (purple). Select transmembrane residues (orange) are shown. Heteroatoms are coloured by type (O, red; N, blue). c The surface of PmtCD coloured by residue conservation as calculated in ChimeraX using a range of −1 to 1. d The surface of PmtCD colored by electrostatic potential as calculated by ChimeraX.

Full size image

To probe if the choice of membrane mimetic affects PmtCD activity, having resolved three distinct states solubilized in peptidisc we first determined the activity of PmtCD in the peptidisc mimetic to have a Vmax of 25.3 ± 1.3 mol Pi/min mol protein (Fig. S9b). Subsequently we compared the relative activity in nanodisc and detergent solubilized PmtCD which revealed slight increases when PmtCD was reconstituted into MSP1D1 nanodiscs and a decrease in activity when extracted in LMNG micelles, presumably due to the replacement of native structural lipids with LMNG detergent (Fig. S9). The addition of PSMɑ3F3A, PSMδ, or LL37 peptide substrates generally decreased ATPase activity in all three mimetics except for PSMɑF3A incubations in the LMNG extracted sample (Fig. S9). These are consistent with our previous results where the presence of PSMs decreased ATPase activity15.

The ADP-bound PmtCD structure represents an intermediate between the ATP-bound closed and nucleotide-free open states

In addition, we have determined the structure of ADP-bound PmtCD in peptidisc at 4.3 Å resolution (Fig. S2). The ADP-bound state of PmtCD, representative of a post-ATP hydrolysis product state, adopts a closed conformation similar to that observed in the ATPγS-bound structure of PmtCD in peptidisc15 (Fig. 4, S10 and S11). The interactions between the TMDs of PmtD are virtually identical with interface areas of ~1140 Å2 (ADP-bound) and ~1190 Å2 (ATPγS-bound) (Cα RMSD ~ 1.1 Å across all 247 residues). In this closed conformation, the two opposing TM5 helices pack against one another, collapsing the intramembrane lumen and splitting it into two symmetrically disposed surface cavities each outlined by TM1, TM2, and TM5 of one protomer and TM5 of the neighbouring protomer (Fig. 4a). The latter protomer caps the cavity at the extracellular side with its EHs. At the intracellular side, the cavity is open and could accommodate the acyl tails of lipids that are possibly tethered to the patch of lysine residues at the putative substrate entrance (Fig. 4b).

Fig. 4: Structural features of ADP-bound PmtCD in peptidisc.
figure 4

a Cavities of ADP-bound PmtCD in peptidisc are shown in cyan with PmtC protomers in yellow and orange and PmtD protomers in blue and red. The predicted membrane interfaces are marked by gray disks, and ADP is shown in green. Heteroatoms are coloured by type (O, red; P, orange; N, blue; Mg2+, green). b The cytosol-membrane interface and the intracellular regions of the two TMD-NBD modules outline a cytosolic cavity predicted to be the substrate entrance. The cavity is adjacent to a cluster of positively charged residues (blue) on the NBD and a patch of invariant polar residues at the cytosol-membrane interface (purple). Select transmembrane residues (orange) are shown. c The nucleotide-binding site shown with one PmtC protomer in yellow and the other in orange. Bonds between ADP and PmtCD are indicated by dashed lines. Region of the map around ADP shown as grey mesh at a contour level of 0.0229.

Full size image

Unlike the TMDs, the NBD dimerization interface in the ADP-bound structure is less compact than the ATPγS-bound state (interface area ~1460 Å2 vs ~1925 Å2, respectively) and represents an intermediate closer in structure to the nucleotide-free open state (interface area ~1380 Å2). Like ATPγS, each ADP molecule is sandwiched between the Walker A motif (33GKNGVGK39) of one NBD and the 120YSMGM124 motif of the neighbouring NBD (Fig. 4c and S11c); however, somewhat less extensive nucleotide interactions with both protomers are observed (~320 Å2 vs ~250 Å2 and ~190 Å2 vs ~110 Å2, for the ATPγS or ADP-bound structures respectively) where specifically the lack of the γ phosphate group results in the less compact dimer observed.

Comparison of ATP, ADP, and apo states in peptidisc

We have now determined the structures of the nucleotide-free apo, ATP-analog bound pre-hydrolysis, ADP-bound post hydrolysis and dissociation of the terminal inorganic phosphate, in a common peptidisc membrane mimetic (Fig. 2a) and at appropriate resolutions to allow a direct comparison of the conformational changes caused specifically by ATP binding, hydrolysis and release.

In the nucleotide-free state, the PmtCD dimer is in an open conformation (Figs. 2b, 3). The symmetrically disposed PmtD TMDs are completely separated with a large intramembrane lumen that is laterally accessible across the full span of both the inner and outer leaflets of the cytoplasmic membrane (Fig. 3a). The open state of the PmtD TMDs appears to be promoted by the looser self-association of the PmtC NBDs, which are less intimately packed in the absence of nucleotide. The tightly associated PmtC ACT domains likely function here, at least in part, to act as a cytosolic tether to ensure complete dissociation of the complex is avoided. Indeed, residues at the dimerization interface of the PmtCD ACT domain are highly conserved and essential for PSM export13. Upon nucleotide binding, the NBDs clamp around ATP resulting in a more tightly associated PmtC dimer interface (Fig. 2a). Comparison of the structures reveals the hinge point for this motion to be the terminal PH helices of PmtC, which pin the NBDs together at two symmetry-related sites (Fig. 2b, c and S5). The TMD and NBD domains are linked through an interaction network formed between α2 of the NBD and CnH and CpH of the TMD (Fig. S12) and this interface does not significantly change between the different states with an interface area of ~1100 Å2 in each case. As a result, motion in the NBD is transduced to the TMDs with a rigid-body translation/rotation of ~13 Å to bring the two PmtDs in close association via a hydrophobic interface along the length of TM5, in turn effectively closing the intramembrane lumen (Figs. 3 and 4). Notably, all resolved nucleotide bound states of PmtCD adopt a closed state rather than an outward-facing conformation for substrate release as typically observed with ABC transporters that employ an alternating access model of transport21. Hydrolysis of ATP to ADP and dissociation of inorganic phosphate leads to the more relaxed NBD dimer observed here. There is a small concomitant shift observed in the PmtD dimer but the relaxation of the NBDs does not appear to be significant enough to induce separation of the TMDs (Fig. 2c and S10). Subsequent ADP dissociation or substrate interactions presumably further destabilize the NBD dimer to provide sufficient energy to dissociate the hydrophobic interface along PmtD TM5, allowing it to return to the open state with reestablishment of the intramembrane lumen. In all three captured states of PmtCD, the basket domain dimer is similarly positioned and tilted ~25ᵒ to the predicted cytosol-membrane plane (Figs. 3, 4, and S11) implying a role of dimer stabilization in the presence or absence of nucleotide. It is also interesting to note that regardless of the conformational state of PmtCD and the membrane mimetic that is present, the corresponding belts were all found to be ~105 Å in diameter representing a common ordered layer of amphipathic molecules stabilizing the TMDs (Fig. S6).

The open state of PmtCD provides a hydrophobic channel for substrate transport

With these structures defining an ATP-hydrolysis induced conformational cycle of PmtCD now in hand, coupled with recent characterizations of three type V ABC transporters which transport amphipathic (but lipid based) cargo by potentially analogous mechanisms, we sought to better understand to which state membrane intercalating PSM peptide substrates likely bind and how the observed ATP driven PmtCD structural changes effect transport.

Eukaryotic type V ABC transporters ABCA1 (cholesterol export), ABCA3 (surfactant lipid export), and ABCA4 (retinal import) share commonalities with PmtCD in structure, sharing the same nucleotide-free open and ATP-bound closed states (see Supplementary Movie 1 for comparison of PmtCD and ABCA1), and function, transporting hydrophobic substrates across a membrane. Structures of ABCA122, ABCA323 and ABCA424,25 in the nucleotide-free open state reveal transporters with separated TMDs forming large channels of comparable volume to PmtCD of ~10,000 Å3 (Fig. 1, S13a–c and S14). Further, for ABCA1 and ABCA4, substrates have been demonstrated to bind the open state: in ABCA1, a cholesterol molecule binds a hydrophobic pocket comprised of TM1, TM2, TM5 and TM11 while analogous residues in ABCA4 coordinate a phospholipid (substrate NRPE is bound in the same channel but at the opposite leaflet of the membrane) (Fig. S13b, c). In PmtCD, comparable residues are observed along TM1, TM2 and TM5 that could serve as a peptide binding site (Fig. S13d–i). Our previous binding data suggested detergent solubilized PmtCD can bind at least two copies of PSMα3F3A15, a mutant PSM that retains cytotoxicity but does not spontaneously polymerize into large fibrils in vitro, a confounding hurdle generally of in vitro analysis of PSMs at higher concentrations14. The significantly improved resolution of the open-state PmtCD structures here now provides a detailed view of the intramembrane lumen dimensions and the specific amino acids that define the putative substrate interacting surfaces of the channel. The lumen volume is able to sterically accommodate one, or more intimately, two peptides, but is incompatible with the dimensions of the higher order PSM fibrillar form characterized previously in isolation by X-ray crystallography (Fig. 5a, b). The PmtCD transporter lumen surface is virtually exclusively hydrophobic that would be complimentary to the hydrophobic faces of the amphipathic PSMs (Fig. 5b and S7), suggesting the binding of two amphipathic PSMs would allow them to shield their polar faces together within the hydrophobic channel. To explore this further, here we selected four residues for mutagenesis from helices TM1 and TM5 that form the entrance to the common hydrophobic pocket on each PmtD protomer and are either located close to the inner (Leu19 and Met166) or outer (Leu30 and Ile181) leaflets (Fig. 5b). Substitutions to either methionine (maintain hydrophobicity but flexible and extend further into cavity) or tryptophan (maintain hydrophobicity but significantly bulkier) (Fig. 5b) were created, confirmed to have similar expression levels (Fig. S15), and tested for transport using a previously designed bacterial secretion assay15 (see methods). Of these, mutation of Ile181 on TM5 at the outer leaflet to the bulkier tryptophan significantly abrogated efflux (Fig. S15) implicating this region of the hydrophobic cavity in function, either directly by interfering with peptide binding, or indirectly by interfering with the open-closed structural dynamics of the TMDs presumed to be involved in secretion.

Fig. 5: Proposed mechanism of PSM loading by PmtCD.
figure 5

a Dimensions of AlphaFold predicted PSMɑ4 dimer, and x-ray structure (6GQC) of a PSMɑ3 fibril shown in purple cartoon with residues shown as sticks. Heteroatoms are coloured by type (O, red; N, blue; S, yellow). b Nucleotide free PmtCD-peptidisc model shown coloured by hydrophobicity either with a generated surface or cartoon. Hydrophobic residues lining the lumen are shown as sticks and residues chosen for mutational analysis labelled. A PSMɑ4 predicted dimer (purple) is docked into the lumen of PmtCD with residues shown as sticks.

Full size image

Discussion

Staphylococci use PmtCD to preferentially secrete PSMα peptides, which play a variety of pathogenic roles including host cell membrane destruction, biofilm development, and the triggering of inflammatory responses8,9. In this work, we have presented nucleotide-free structures of PmtCD in an open conformation in three different membrane mimetics along with an ADP-bound structure in a closed conformation. Together with our previously reported ATPγS-bound structure15, we have now captured structural snapshots of the nucleotide-free (open), pre-hydrolysis nucleotide-bound (closed) and post-hydrolysis and inorganic phosphate released (relaxed) states. Importantly, we have determined the structures of each in a consistent peptidisc membrane mimetic. Together, these delineate distinct conformational states involved in the essential efflux of membrane-disrupting PSMs.

In lieu of a substrate-bound structure where alternative states may exist, comparison between the eukaryotic ABC transporters ABCA1, ABCA3 and ABCA4, with the open and closed states of PmtCD, supports a similar lateral access and extrusion mechanism whereby substrate binds the open state with separated TMDs and subsequent ATP binding-induced channel closure serves to propel the substrate across the membrane. The amphipathic nature of substrate PSMs could permit partition into the inner leaflet of the bacterial membrane, where peptides would enter the exporter laterally with their nonpolar face presumably buried in the lipid bilayer. Once bound within the hydrophobic lumen of PmtD in the open form, peptides could oligomerize to shield their polar faces from the hydrophobic membrane environment of the inner lumen to facilitate passage. Our structures of the transporter apo form are of a size that could feasibly accommodate a dimer pair of helical peptides. Alternatively, if exported in a monomeric form, the exposed hydrophilic face could potentially destabilize the packing within the hydrophobic lumen and help energize expulsion. Subsequent ATP binding would induce association of the PmtD protomers with collapse of the intramembrane lumen, leading to transport of the substrate across the membrane. Despite extensive trials, we have been unable to capture PmtCD in complex with a PSM peptide substrate using cryo-EM, leaving open the possibility of intermediate states that may help answer critical questions such as what prevents peptides from exiting laterally back into the membrane or if alternate substrate entry points exist. Together with the difficulty in obtaining clean binding data by the many methods tried (MST, ITC, SPR) and the preference of PmtCD to act on a variety of the most hydrophobic of the PSMs, we suggest a largely hydrophobic export trajectory where capture of a single defined substrate complex of sufficient occupancy to allow high resolution cryo-EM reconstructions may be a significant challenge.

Lipid molecules that surround PmtCD are also likely to play a critical role in PSM transport. In the open conformation of PmtCD, we predict lipids occupy the intramembrane lumen, perhaps using the significant cluster of basic residues at the cytosol-membrane interface as a tether as has been observed in structures of other ABCA transporters (Fig. S13b, c). In addition, the reentrant helices of PmtD form a cap at the extracellular side of the membrane which could displace lipid headgroups and create a region that is loosely occupied by the acyl tails of surrounding lipid molecules. These reentrant helices are a characteristic feature of type V ABC transporters16 and have been implicated in creating a favorable environment for lipid headgroups to flip across the membrane in ABCA426. It is tempting to speculate that the formation of a sparsely populated lipidic region below the reentrant helices provides the space required for rapid displacement of lipids from the intramembrane lumen when the transporter closes.

In further support of PmtCD transporting membrane partitioned substrates, an additional S. aureus ABC transporter – AbcA (distinct from the eukaryotic ABCA transporters) – has been shown to also export PSMs, cooperating with PmtCD to either export substrates from the cytosolic (AbcA) or membrane (PmtCD) environments27. It was found that the hydrophobicity of the PSMα peptide influenced which transporter was preferred with more hydrophobic peptides that are more likely to be partitioned into the membrane e.g., PSMα4, preferentially exported by PmtCD and more hydrophilic peptides e.g., PSMα3, exported by AbcA. Of note, only PmtCD provides self-immunity against PSMα peptides, presumably by preventing the toxic accumulation of PSMs that partition into the S. aureus bacterial membrane27.

Finally, PmtCD is also responsible for protection against mammalian antimicrobial peptides (AMPs), including the cationic amphipathic peptide LL-3712. How these bind PmtCD is uncertain given they are assumed to accumulate in the outer leaflet of the membrane; however, by analogy with NRPE binding and translocation by ABCA4 at the equivalent membrane leaflet (e.g., on the opposite side to the NBDs), it is possible that PmtCD could sequester AMPs similarly from the outer leaflet and eject them out of the membrane upon nucleotide binding and domain closure. Thus, given the essential role of PmtCD in both self-immunity and protection against host defence reinforce, the inhibition of PSM secretion through therapeutic intervention provides a logical trojan horse target for treatment of the increasingly drug resistant infections of this pathogen.

Materials and methods

Expression and purification

S. aureus RN4220 cells, transformed with the pTX17-PmtCDhis expression plasmid15, were grown at 37 °C for 20 hours in tryptic soy broth (TSB) without dextrose medium supplemented with 12.5 μg/ml tetracycline and 0.5% D-xylose. The cells were harvested by centrifugation and resuspended in phosphate-buffered saline (PBS) with 10 mg/ml deoxyribonuclease (DNase) I and 1 mg/ml lysostaphin. After a 30 min incubation at 37°C, the cells were placed in an ice water bath and sonicated at 50% power for a total of 6 min. Cell debris was pelleted by centrifugation at 15,000 × g for 30 min and the supernatant was collected for centrifugation at 200,000 × g for 45 min. The membrane pellet was resuspended in buffer A (20 mM Tris pH 8.8, 200 mM NaCl, 5% glycerol, 20 mM imidazole) and either 1% DDM or LMNG was added for PmtCD solubilization.

To prepare LMNG-solubilized PmtCD for cryo-EM analysis, resuspended membranes were incubated with 1% LMNG for 1 hour at 4°C. Solubilized proteins were then separated from the remaining membrane by centrifugation (200,000 × g for 30 min at 4 °C) and incubated with Ni-NTA (Thermo Scientific). The resin was washed with buffer A supplemented with 0.003% LMNG. PmtCD was eluted with buffer B (20 mM Tris pH 8.8, 200 mM NaCl, 5% glycerol, 300 mM imidazole, 0.003% LMNG) and further purified by size-exclusion chromatography using a Superose Increase 6 10/30 column (GE Healthcare Biosciences) pre-equilibrated with buffer C (20 mM Tris pH 8.8, 200 mM NaCl, 0.003% LMNG). PmtCD was reconstituted into peptidiscs and nanodiscs using DDM-solubilized protein. Extraction and purification of PmtCD in DDM and reconstitution of PmtCD into peptidisc were performed as previously described in ref.15. Briefly, for reconstitution of PmtCD into MSP1D1 nanodiscs: PmtCD extracted in 1% DDM and affinity purified via Ni-NTA was mixed overnight in a 0.5:1:50 ratio of PmtCD:MSP1D1:lipid mixture (70% DMPG, 30% cardiolipin) in the presence of biobeads to remove detergent. Reconstituted sample was centrifuged for 20 min at 57,000 x g in a TLA 110 rotor and further purified via gel filtration on a Superose Increase 6 10/30 column (GE Healthcare Biosciences) pre-equilibrated with buffer D (20 mM Tris pH 8.8, 200 mM NaCl).

Cryo-EM sample preparation and data collection

Three-microliter aliquots of samples at 1 mg/ml (in either LMNG, peptidiscs, or nanodiscs) were applied to glow-discharged Quantifoil grids (Copper, 300 mesh, R2/2), blotted for 3 s at 100% humidity and plunge-frozen into liquid ethane using a Vitrobot Mark IV (Thermo Fisher). An ADP-bound PmtCD sample was prepared by incubating PmtCD reconstituted into nanodiscs with 5 mM ADP for 1 h at 4 °C. Grids were screened at the High Resolution Macromolecular Cryo-electron Microscopy (HRMEM) facility (Vancouver, British Columbia, Canada) on a 200 kV Glacios microscope (Thermo Fisher) equipped with a Falcon III camera (Thermo Fisher).

For PmtCD reconstituted into nanodiscs, a total of 2508 movies were collected on a 300 keV Titan Krios microscope (Thermo Fisher) located at the HHMI Janelia Research Campus equipped with a spherical aberration corrector, an energy filter (Gatan GIF Quantum) and a post-GIF Gatan K3 Summit direct electron detector. Images were taken on the K3 camera in dose-fractionation, corresponding to 1.30 Å per physical pixel (0.65 Å per super-resolution pixel). Each movie series contained 40 frames and each frame received a dose of 1.079 electrons per Å2. An energy slit with a width of 20 eV was used during data collection. Fully automated data collection was carried out using SerialEM with a nominal defocus range set from −1.5 to −3.5 μm. For PmtCD reconstituted into peptidiscs and incubated with ADP, a total of 2948 movies were collected on a 300 keV Titan Krios microscope located at the HHMI Janelia Research Campus equipped with a K3 camera in dose-fractionation, corresponding to 1.35 Å per physical pixel (0.675 Å per super-resolution pixel). Each movie series contained 50 frames and each frame received a dose of 1.16 electrons per Å2. For PmtCD in LMNG, a total of 1074 movies were collected on a 300 keV Titan Krios microscope located at the HRMEM facility equipped with a Falcon III camera in counting mode, corresponding to 1.09 Å per physical pixel. Each movie series contained 48 frames and each frame received a dose of 1.25 electrons per Å2.

Cryo-EM data processing and model building

Unless otherwise specified below, cryo-EM data were preprocessed in RELION-3.128. Motion correction was performed using MotionCor229, and contrast transfer functions of the summed and dose weighted micrographs were determined using CTFFIND430. Approximately 2000 particles were manually boxed out from selected micrographs to generate reference-free 2D-class averages. The representative 2D-class averages were then used as templates for automated particle picking.

For the PmtCD in LMNG dataset, 489729 particles were picked in RELION-3.1 and transferred to cryoSPARC v3.231 for reference-free 2D classification. 119004 particles from select 2D class averages were then used to generate two initial volumes. After multiple rounds of heterogeneous refinement, the best volume was selected for homogeneous refinement with C1 symmetry. The resulting map was used as an initial model for a second round of 2D classification starting with the initial group of 489729 particles. 143119 particles, generating an improved map, were transferred back to RELION-3.1, and two rounds of 3D classification were performed. The particles from the best volumes were grouped and used for Bayesian polishing, CTF refinement, and partial signal subtraction to remove the micelle density. One round of 3D classification without alignment was performed and 95729 particles belonging to the best volumes were grouped and refined one last time. The resulting volume with C1 symmetry is at 3.8 Å resolution. A map at 3.6 Å resolution with C2 symmetry was then generated after signal subtraction of the basket domain.

For the PmtCD reconstituted into nanodiscs dataset, movies were binned to 1.30 Å per pixel for processing. Motion correction, CTF estimation and particle picking were performed as outlined above. With 614993 particles, four initial volumes were generated and subjected to multiple rounds of heterogeneous refinement before the best volume was selected for homogeneous refinement with C1 symmetry. The refined map generated with 108661 particles were transferred back to RELION-3.1 for Bayesian polishing, CTF refinement, and partial signal subtraction to remove the peptidisc density. The resulting volume with C1 symmetry is at 4.3 Å resolution. A map at 4.2 Å resolution with C2 symmetry was then generated after signal subtraction of the basket domain.

For the dataset on PmtCD reconstituted into peptidiscs and incubated with ADP, movies were binned to 1.35 Å per pixel for processing. Motion correction, CTF estimation, and particle picking were performed as outlined above. 1580766 particles were picked in RELION-3.1 and transferred to cryoSPARC v3.2 for reference-free 2D classification. 158239 particles from select 2D class averages were then used to generate two initial volumes that revealed the presence of two distinct conformational states: an open state and a closed state. After multiple rounds of heterogeneous refinement, the best volume for each of the two conformational states was selected for homogeneous refinement with C1 symmetry. The two resulting maps were used as initial models for a second round of 2D classification starting with the initial group of 1580766 particles. 147138 particles of PmtCD in the open state and 142155 particles of PmtCD the closed state, generating improved maps, were separately transferred back to RELION-3.1 and subjected to multiple rounds of 3D classification. For each conformational state, the particles from the best volumes were grouped and used for Bayesian polishing, CTF refinement, and partial signal subtraction to remove the nanodisc density. One round of 3D classification without alignment was performed and particles belonging to the best volumes were grouped and refined one last time for each of the two states (71603 particles for the open state and 63048 particles for the closed state). The two resulting volumes with C1 symmetry are both at 4.3 Å resolution. Two maps with C2 symmetry were then generated after signal subtraction of the basket domain both at 4.2 Å resolution.

Resolutions were determined according to the gold-standard Fourier shell correlation (FSC) 0.143 criterion. All final maps were sharpened using deepEMhancer32. For model building, the four domains of PmtCD from PDB 6XJI were first individually docked into each cryo-EM map using UCSF Chimera33. The structures were then refined using Phenix34 and Coot35.

Analysis software

Structures were visualized and represented using UCSF ChimeraX36, UCSF Chimera33 and PyMOL37. Sequence alignments were produced using the ESPript 3.0 server38. For sequence conservation analyses, sequences of PmtC and PmtD homologs were acquired from Uniprot39 and assessed with UCSF ChimeraX36. The positioning of proteins in membranes were predicted using the PPM Web Server40. Lumens and cavities were identified with the solvent extractor tool (inner probe radius of 4 Å and outer probe radius of 10 Å) of the Voss Volume Voxelator server19. Visual figures were created with GraphPad Prism 10, and Adobe Illustrator 2025 (Adobe).

PmtCD ATPase activity

ATPase activity of PmtCD was measure using a malachite green assay similarly to other type V ABC transporters24,41. ATPase assays for Vmax determination were performed in 30 μL reactions shielded from light in 96 well flat-bottom NUNC trays with SEC purified PmtCD diluted to 1 μM in buffer C supplemented with ATP ranging from 4 mM to 1.9 μM, and incubated at 30 °C for 15 min. After incubation at 30 °C developing solution (0.9 mM Malachite green, 9 mM ammonium molybdate, and 0.1% (v/v) triton X-100) was added to the mixture in a 1:5 (reaction to developing solution) ratio to dye free phosphate and incubated for another 15 min at RT. Substrate treated reaction mixtures were composed of 1 uM PmtCD, 20 uM PSMɑF3A, PSMδ, or LL37, 250 uM ATP, and 125 uM MgCl2. PmtCD samples were pre-incubated for 30 min at RT with substrates followed by addition of ATP and MgCl2 and further incubation at 30 °C for 30 min and developed for 30 min at RT for peptidisc and nanodisc reconstituted samples. LMNG extracted samples were incubated at 30 °C for 60 min and developed for 60 min at RT. ATPase activity was quenched using 34% (w/v) citric acid and absorbance read at 595 nm using a SYNERGY H1 microplate reader. Free phosphate concentration was estimated based upon K2HPO4 standards, and statistical analysis performed using GraphPad Prism 10.4.0. Background adjusted 595 nm absorbances and conversion to mol Pi/min*mol PmtCD are reported in supplemental data 1.

PmtCD mutagenesis

Mutations to pRB473 pmtCD were engineered as before15 using inverse PCR to introduce codon changes. PCR products were treated with T4 PNK and T4 ligase before being electroporated into S. aureus RN4220. All mutants were confirmed by PCR and Sanger sequencing.

PSM secretion assay

Secretion of PSMs were measured as before15. Briefly, S. aureus LAC ΔPSMs Δpmt were co-transformed with plasmids (Table S2) encoding the indicated pmtCD mutant and the PSM α1-4 operon under control of a xylose inducible promoter. PSM expression was induced for 1 hour, after which culture supernatants were collected and analyzed by HPLC-MS (Agilent 1260 LC, Agilent 6120 quadrupole) in selected ion monitoring mode to quantify secretion of each. A linear model was used to compare relative PSM secretion for each amino acid substitution with wild-type PmtCD, controlling for PSM identity and HPLC injection order. Peak area and normalized PSM ɑ1-4 secretion percents for each mutant are reported in supplemental data 1.

PmtD Western blot

Equal amounts by OD600 of S. aureus cells immediately after the 1-hour induction were lysed with glass beads and solubilized with SDS sample buffer. The samples were separated by SDS-PAGE using a 10% acrylamide gel and transferred to a nitrocellulose membrane. Western blots were performed using the cell culture supernatants of the hydridoma clone 2D142. The blots were imaged with an Amersham Typhoon 5 scanner (Cytivia).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.